Disable health on single-rail deployments, since the concept of resiliency doesn't exist on single-rail deployments. There is no other interface to fail over to.

local failures:

• 1. Don't call lnet_handle_local_failure() if I only have a single rail.
• 2. Don't queue for resend if I only have a single rail.
  remote failures:
• 1. Don't call lnet_handle_remote_failure() if the destination only has a single rail.
• 2. Don't queue for resend if the destination only has a single rail.